diagnose: Diagnose a learner, strategy, or model with respect to a...
In Prometheus77/ezmlr: Easy Interface to mlr

Description Usage Arguments

View source: R/diagnose.R

Given a task and either a strategy or learner, or else given a trained model, create diagnostic plots to provide useful information about model performance and the inner workings of the model.

diagnose(obj, task, plots = NULL, threshVsPerf = NULL, ROCCurve = NULL,
  residuals = NULL, learningCurve = NULL, filterValues = NULL,
  hyperParsEffect = NULL, optPath = NULL, tuneMultiCritResult = NULL,
  PartialDependence = NULL, calibration = NULL, ...)

`obj`	(`learner`, `strategy`, or `model`) The object to diagnose
`task`	(`task`) If `obj` is a `learner` or a `strategy`, a `task` must be specified on which to evaluate performance.
`plots`	(`list`) List of plots to produce. If not specified, the function will choose a set of appropriate plots based on the object and task. Will be overridden if any of the individual plot parameters is set to `TRUE` or `FALSE`.
`threshVsPerf`	(`logical`) Plot threshold vs. performance curve using `plotThreshVsPerf`? Only available for classification tasks.
`ROCCurve`	(`logical`) Plot receiver operating characteristic (ROC) curve using `plotROCCurves`? Only available for classification tasks.
`residuals`	(`logical`) Plot residuals of predictions using `plotResiduals`?
`learningCurve`	(`logical`) Plot learning curve showing learner performance vs. proportion of data used, using `plotLearningCurve`?
`filterValues`	(`logical`) Plot feature importance for a given filter method using `plotFilterValues`?
`hyperParsEffect`	(`logical`) Plot hyperparameter effect on model performance from tuning data using `plotHyperParsEffect`?
`optPath`	(`logical`) Plot optimization path from tuning data using `plotOptPath`?
`tuneMultiCritResult`	(`logical`) Plot pareto front for results of tuning to multiple performance measures using `plotTuneMultiCritResult`?
`calibration`	(`logical`) Plot calibration of probability predictions vs. true incidence using `plotCalibration`?
`...`	Additional arguments to pass to each plot.
`partialDependence`	(`logical`) Plot partial dependence of model prediction over each data feature using `plotPartialDependence`?